Equinix Nearly Doubles Capex Outlook as AI Inference Demand Accelerates

Why Equinix Is Nearly Doubling Its Data Center Spending Plans

Equinix, the world's largest data center company, used its second-quarter results to make an unusually forceful bet on the AI infrastructure boom: it raised its revenue outlook for the rest of the decade and said annual development spending could now more than double the level it forecast just a year ago.

The company reported revenue up 16% year over year in the second quarter, with bookings growing more than 30% and the bulk of its largest deals tied to artificial intelligence workloads. But the more striking news came from forward guidance. Equinix now expects 2026 revenue growth of 12%, roughly $100 million above its prior projection, and sees annual growth of 10% to 13% between 2027 and 2029 — well above the 7% to 10% range it gave a year earlier. CEO Adaire Fox-Martin called it “the largest single guidance raise in the history of our company.”

The spending commitments behind those numbers are just as dramatic. Planned capital expenditure for 2026 jumped from $3.8 billion to as much as $6 billion, and Equinix now projects $5 billion to $7 billion of annual capex from 2027 through 2029, up from a prior $3 billion to $4 billion range. Management said the increase reflects both stronger-than-expected demand in the latest quarter and growing conviction that the AI market is evolving toward Equinix's model: smaller, connectivity-heavy data centers near population centers rather than the giant remote campuses favored by hyperscale cloud providers.

The company argues that as AI shifts from training models to inference — running those models in real-world applications — workloads will increasingly need to sit close to end users and connect easily across multiple clouds and networks. That, executives say, is exactly what its interconnection-focused portfolio delivers. Around 80% of the new development spending will go into Equinix's top 25 markets, where demand already exceeds supply and new construction faces significant power and permitting hurdles. The company controls 3 gigawatts of developable land, with roughly 770 megawatts currently under development. Wall Street responded approvingly, lifting the REIT's shares about 4% by midday Thursday, though analysts also pressed management on what happens if demand does not materialize as planned.

The Inference Bet: Why AI Demand Is Moving Toward Equinix's Model

The Scale of the Raise, and What It Signals

Guidance moves of this size are rare for a company of Equinix's maturity. Raising the 2027-2029 revenue growth band by three percentage points and lifting 2026 capex by more than half — from $3.8 billion to as much as $6 billion — is a commitment made before the revenue it targets is locked in. CFO Olivier Leonetti said most of the capital will go to markets Equinix knows well “where demand is higher than supply,” which is management's way of saying the company is pricing in scarcity rather than speculative growth. Investors appear to accept that logic for now, but the guidance raise also resets expectations: the market will measure Equinix against the new, higher numbers from this quarter forward.

Why the Shift to Inference Favors Interconnection

Equinix's core bet is a thesis about how AI computing evolves. Training massive models works fine in huge, cheaply powered campuses far from cities; running those models in everyday applications — a chatbot for a bank's customers, a logistics tool for a retailer — requires low latency and easy access to multiple clouds. Equinix's portfolio of interconnection hubs in dense metro markets is built for exactly that second stage. The company says the shift is arriving faster than expected, with Fox-Martin citing “acceleration in the AI infrastructure cycle” that plays to its strengths. This remains a claim about the future, not a fact of the present: bookings growth is strong, but the inference-heavy demand mix Equinix is banking on will only be proven out over the next several quarters as enterprise AI deployments scale.

Spending Where Power Is Scarrest

The geographic concentration of the plan is telling. By directing roughly 80% of development spending to its top 25 markets, Equinix is deliberately avoiding the land-banking race in less constrained regions and betting instead on its ability to deliver capacity where utilities, contractors and community relationships are already in place. The 3 gigawatts of developable land under control — with about 770 megawatts in active development — gives the company an unusual long-dated inventory in markets where new entrants would need years to secure power. That is the competitive moat, but it is also the execution risk: much of what Equinix builds depends on utility interconnection timelines and local approvals it does not control.

The Overbuilding Risk No One Can Rule Out

Equinix is not alone in making oversized commitments to AI infrastructure, and the industrywide debate about whether those commitments will pay off has already rattled even the biggest hyperscalers. Morningstar analyst Martin Szumski captured the concern in a Thursday note: “The scale of this infrastructure buildout raises concerns that data center operators may be left with excess capacity and poor returns on capital.” Equinix's defense is that its facilities are differentiated and its markets supply-constrained. The counterargument is that demand projections of 10% to 13% annual revenue growth through 2029 compound on a very large base, and any slowdown in enterprise AI adoption would leave utilization — and return on that $5 billion to $7 billion of annual capex — below plan.

What to Watch After Equinix's Record Guidance Raise

For investors and data center customers, the concrete things to track are the ones Equinix itself has now put a number on:

• Booking momentum: bookings grew more than 30% year over year in Q2, with AI driving most large deals. A deceleration here would be the earliest sign that the $6 billion 2026 capex plan is outrunning actual demand.
• The 2027-2029 revenue band: Equinix has committed to 10% to 13% annual growth against last year's 7% to 10% range. Any future narrowing toward the bottom of that band would signal the inference thesis is weaker than the CEO's “largest single guidance raise in the history of our company” implies.
• Execution in the top 25 markets: with 80% of development spending concentrated there and 770 megawatts already under construction from a 3-gigawatt land pipeline, Equinix's results increasingly hinge on utility power delivery and permitting timelines it does not control.
• Capacity in interconnection-dense metros: enterprises shifting AI workloads to inference should expect premium pricing and limited availability in Equinix's strongest markets for the foreseeable future — the guidance raise is, in effect, a signal that supply there will stay tight even with record spending.
• Industry capex signals: if large hyperscalers' public concerns about AI spending — the same worries that have hit the likes of Google — turn into actual cutbacks, data center demand broadly would soften, and Equinix's interconnect-heavy niche would not be insulated.